qhjqhj00 otb99-eval
ReasoningTrack: Chain-of-Thought Reasoning for Long-term Vision-Language Tracking — Xiao Wang et al. (2025) (arXiv:2508.05221, 2025)
What this evaluates
Evaluates short-term vision-language tracking performance on a curated subset of OTB100 with added textual annotations, testing robustness to appearance changes and scale variations.
Datasets
- OTB99 — total 99; splits: train (51), test (48)
Metrics
PR(primary) — range: percent- Precision Rate: percentage of frames where the center distance is within 20 pixels.
AUC— range: percent- Area Under the Curve: area under the precision plot across varying distance thresholds.
Input / output format
Input: Video frames, initial language description, and ground-truth bounding boxes.
Output: Predicted bounding box coordinates per frame.
Scoring recipe
For each frame, compute center distance d.
PR = (count(d < 20) / N) * 100
AUC = integral of precision plot over thresholds [0, 50]
Common pitfalls
- Text annotations are added post-hoc to OTB100, so baseline methods may not be trained with language guidance.
- Short-term tracking may not stress-test long-horizon language updates.
Evidence (verbatim from paper)
OTB99 consists of 99 video sequences, divided into 51 for training and 48 for testing. As shown in Table[V], our method achieves the best performance in two metrics, with an AUC of 71.11% and PR of 95.58%.
Citation
@misc{wang2025reasoningtrack,
title={ReasoningTrack: Chain-of-Thought Reasoning for Long-term Vision-Language Tracking},
author={Xiao Wang et al. (2025)},
year={2025},
note={arXiv:2508.05221}
}
- arXiv: 2508.05221